Sentinel mechanism of redis cluster

When redis Cluster master （master） It's down. , The cluster can only read but not write ！ How to solve this problem ？

The answer is ： Sentinel mechanism , Solve the problem of master-slave replication The election question

One Sentry group set up

1. Server introduction

Here's the picture ,redis The cluster of 5 Servers , Master node ip by 192.168.1.100

redis For cluster setup, please refer to ： redis One master multi slave cluster setup _huangyuewang The column -CSDN Blog

sentry Mechanism principle

1. Sentinel mechanism, each 10s We only need to configure our master node to monitor the time, and then we can get the current whole time Redis List of cluster environments , use info  Order form .
2. Sentinels are not recommended to be stand-alone , It's better for each one Redis All nodes need to be configured with sentinel monitoring .
3. What is the principle of sentinel cluster ： Multiple sentinels perform the same master master node , Subscribe to the same channel , When new sentries join, they will send their own service information to the channel , Subscribers to the channel can find the addition of new sentinels , Then establish a long connection with each other .
4. Master Fault discovery of A single sentinel will turn to the Lord master The node sends ping The order of , If master The node did not respond in time , The Sentry will think it's time to master The node is “ Subjective unavailability ” It will be sent to other sentinels to confirm the Master Whether the node is unavailable , Number of sentinel nodes currently confirmed >=quorum( Configurable ), There will be a re-election .

2. Build sentry clusters

 

find redis Unzip folder , hold sentinel.conf File copy to redis The installation directory

cp /application/redis-6.2.6/sentinel.conf /application/redis/etc

  Modify the sentry profile  vim /application/redis/sentinel.conf

  Activate the sentry

/application/redis/bin/redis-sentinel /application/redis/etc/sentinel.conf

Other commands

sentinel down-after-milliseconds mymaster 3000#sentinel Heartbeat detection master 3 No response in seconds , Regarded as hanging up , Start switching from other to main 
sentinel parallel-syncs mymaster 1# There can be at most 1 Slave synchronization master . One from the end of synchronization , The other synchronizes from .
sentinel failover-timeout mymaster 18000# Master slave switching timeout 
 Activate the sentry ./redis-sentinel ./sentinel.conf

Above 5 All servers set up and start sentries